OpenAI, Part 3: The (Probabilistic) Future of OpenAI

Thoughts on the future of the AI company

In Part 1 of my OpenAI deep dive, I said how important OpenAI is. In Part 2, I said they might be as complex as they are important. For Part 3, I’ll add to that: they might be as interesting as they are complex!

What makes investing so interesting is the thought experiment of it all. We get to think about the future, bet accordingly, and (hopefully) make money from it.

For a company like OpenAI, that thought experiment is as interesting as it gets.

Now, if you’re looking for me to make definitive predictions about OpenAI, you’re in the wrong place. A reminder of Uncle Charlie’s quote on thinking probabilistically:

“If you don’t get this elementary, but mildly unnatural, mathematics of elementary probability into your repertoire, then you go through a long life like a one-legged man in an ass-kicking contest. You’re giving a huge advantage to everybody else. One of the advantages of a fellow like [Warren] Buffett, whom I’ve worked with all these years, is that he automatically thinks in terms of decision trees and the elementary math of permutations and combinations.”

So, instead of writing “The Future of OpenAI” piece. I’ll write “The Probabilistic Future of OpenAI” piece. To do that, I have to lay out the variables on which OpenAI’s future MOST depends (I might miss many of them; these are just my thoughts, after all).

Those questions (which should be thought of as spectrums) are:

1. Cost structure: How successful will OpenAI’s pursuit of vertical integration and ensuring cost structure improvements be?

2. Business model: How does OpenAI profit from their models and how commoditized do those models become?

3. Market: What’s the end state of the AI application market (size, market share, profitability)?

4. Product: How do you take the intelligence of models and make the actions of agents a reality?

5. Valuation: How do you value OpenAI?

The first of those variables is vertical integration:

1. Cost Structure: Own it all or die trying!

(Of course, they won’t die trying; they’ve raised tens of billions of dollars, might as well try to own it all anyway!)

There’s more money flowing into AI than any other industry in its early stages in history (estimated $450B in VC funding in the last 4 years, compared to $256B on the dot-com bubble).

That means there’s more capital funding competitors than any other industry in history.

Building competitive advantages in that environment is challenging; to do so, OpenAI is pursuing vertical integration both up and down the AI stack.

As we’ve seen with Stargate and OpenAI’s chip efforts, moving down the stack gives them a cost advantage and the control they need to compete at the hardware level.

Most of their competitors have advantages on either custom hardware (Google, Amazon, Meta) or data centers (xAI and the hyperscalers). OpenAI can’t afford to lose ground to competition because they don’t have enough control over their hardware.

However, OpenAI can survive without its own data centers or hardware. Leading with their applications? That’s an imperative, and it’s the path to sustainable economics.

2. Business Model: How do you create sustainable economics?

For the foreseeable future, billions of dollars will flow into direct competitors (see: Anthropic, xAI, and SSI (maybe a direct competitor)).

On top of that, the largest and most profitable tech companies to ever exist are in an existential arms race to compete with OpenAI.

If that’s not enough, DeepSeek showed you can create a model that’s cheaper to train, cheaper to run, and nearly the same quality.

Some of those competitors will also open-source their models, giving the model away for the price of the cheapest inference provider.

Because of these variables, competing with an API at the model level will not lead to a sustainable business model, at least for the foreseeable future.

That means you need to build software moats, like switching costs and enterprise relationships. The way to do that is at the application layer.

3. Market: What does the AI app market look like in a decade?

If it’s true that OpenAI is a vertically integrated AI company that will make the majority of its profits from the application layer, then this may be the most important question: how big is the AI application market in the long-term, and where will value accrue in that market?

If you measure the AI app market by the potential knowledge work it could replace, the market’s…unbelievably large. However, the price of goods trends towards its marginal cost to produce. That means cost of hardware + cost of energy + cost of AI researchers + whatever margin you can protect at the application layer.

The better way to view this is to ask: What problems will AI be able to solve?

The answer is many of them, and the market will be very large, likely in the hundreds of billions of dollars.

It will be large enough that if OpenAI executes and defends its competitive advantages, it will be much larger than it is today.

On AI app market dynamics, it seems OpenAI sees the potential to dominate general-purpose workflows with their agents (see demo of Deep Research AI for sales workflows.)

If they’re successful, they’ll own a large portion of the general workflows that get passed off to AI.

In markets like healthcare, finance, and legal, vertical-focused solutions likely arise to handle those markets' specific integrations, regulations, and workflows (much like how the software market developed!)

Now, if OpenAI is really successful, they’ll also be the foundation for these vertical companies; they may do so through exclusive access to their leading-edge models, something akin to the luxury handbag market.

What about the timing of the market?

Amara’s Law says, “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

It’s pretty clear that knowledge work in its current form is going to change drastically a decade from now. Will it change drastically a year from now? Who knows.

4. Product: OpenAI has aggregated information; now, how do they aggregate actions?

The value of AI thus far has been in its ability to aggregate information. The ability of AI moving forward is in its ability to take action.

OpenAI knew this many months ago and released Plugins with this vision. Companies could pre-define workflows and authentication for ChatGPT to access. It SHOULD HAVE worked.

But it didn’t. My hypothesis is that the ecosystem wasn’t ready.

Plugins might not have worked because the infrastructure (APIs, integrations, authentication) wasn’t in place for ChatGPT to effectively access websites.

Operator, while less efficient, bypasses that. The logic is simpler now:

IF api == 1, use api.

IF api == 0, bypass it and use Operator.

The next step necessary to develop actions was reasoning: weighing the various variables it takes to make decisions (think of all the micro-decisions when planning travel). OpenAI developed that!

OpenAI has now developed Operator, reasoning models, and agents.

The Future of OpenAI is intelligence, but the future of ChatGPT is actions.

This is the vision for how it gets to $100B in revenue. Not a chatbot but a general-purpose assistant; and eventually, a general-purpose knowledge worker.

5. Valuation: How do you value OpenAI?

This question is fascinating because there are so many variables that investors could be weighing to invest in OpenAI:

1. OpenAI is the clear power law winner of AI; as we’ve seen in past cycles, a large % of value accrues to the power law winner.

2. The opportunity for AI is as big as the hype suggests, and OpenAI is an index on that opportunity.

3. Some form of AGI is coming, and you want a seat at the table when it comes.

4. The information from having access to OpenAI gives you such an edge in rest of the market that it’s worth it.

5. The $157B valuation is a ~39x revenue multiple on this year's revenue. Next year's revenue is projecting around $12B (who knows what that actually ends up being), so that’s closer to 13x forward revenue. That’s expensive but not crazy in public markets. (If the $300B round closes, then these valuation metrics double, which somewhat mutes this point).

In reality, the decision-making process is likely a combination of those variables.

What I know is this: With fast-growing companies, the end state matters much more than any intermediate metrics. You can justify whatever valuation you want based on your pre-conceived biases. If you believe in the long-term value of AI, you can justify the valuations today.

I suspect many of the OpenAI investors are thinking along these lines.

I will leave you with one quote, “If you aren’t willing to own a stock for 10 years, don’t even think about owning it for 10 minutes.”

In times like these, it’s worth reading that twice.

My final thought on OpenAI (for now):

My hope is that this article is something like a time-stamped journal entry on my thoughts on this company. Perhaps OpenAI is a multi-trillion dollar company in a decade. Perhaps it stagnates for many years as it “grows into its valuation” while AI value catches up to expectations.

Perhaps it’s a cautionary tale on early assumptions of market dominance (I’m most skeptical of this outcome).

Because of the Power Law, technology can’t help but create iconic companies. Rapidly growing giants, pushing forward technology, with the biggest brands in the world.

IBM. Intel. Apple. Microsoft. Nvidia. Amazon. Google. SpaceX. Tesla. Google. Meta.

OpenAI has asserted itself into this category of company. The early, dominant player in a rapidly growing market that’s synonymous with the technology it brought forth.

We have the privilege of seeing it in its early stages.

As always, thanks for reading!

Disclaimer: The information contained in this article is not investment advice and should not be used as such. Investors should do their own due diligence before investing in any securities discussed in this article. While I strive for accuracy, I can’t guarantee the accuracy or reliability of this information. This article is based on my opinions and should be considered as such, not a point of fact. Views expressed in posts and other content linked on this website or posted to social media and other platforms are my own and are not the views of Felicis Ventures Management Company, LLC.